The AI Mention Effect: Why Australian Brands Are Getting Traffic They Can't See (And What To Do About It)

We've spent much of the last two years telling clients an uncomfortable truth. The most valuable thing AI does for your brand is also the hardest thing to measure. Someone asks ChatGPT for the best option in your category. It names you. They nod, close the tab, and turn up on your website two days later having typed your name straight into the browser. Your reporting logs a "direct" visit with no source. The AI did the work. Your dashboard took the credit and handed it to nobody.

For a long time this was a hunch, a pattern we kept seeing in client data but couldn't prove. Now there's a study that measures it head on, and it confirms what we suspected was happening under the surface. Here's what it found, why it matters more in Australia than almost anywhere else, and what we tell brands to do about it.

What Profound Actually Measured

The research comes from The AI Mention Effect, published on 1 July 2026 by Nikolas Laskaris, AI Strategist at Profound. Rather than guess at AI's downstream impact, Profound joined two data streams from a privacy-safe, double-opt-in panel: what people saw in AI conversations, and what those same people did next in their web browsing. According to Profound, the study drew on more than two million AI conversations across ChatGPT, Gemini and Google AI Overviews between January and June 2026.

The headline finding is the one every marketer should sit with. According to Profound, after an AI assistant introduced a brand into a conversation, a brand the user had not named themselves, that user went on to visit the brand's website at roughly 1.5 to 2.5 times their forecast baseline rate over the next seven days. Profound isolated this by comparing each person's behaviour after the mention against their own behaviour toward the same brand in the weeks before it, so the lift reflects the mention itself rather than pre-existing interest.

The platform differences are just as instructive. According to Profound, Gemini produced the largest relative lift, around 2.5x its baseline, while Google AI Overviews carried the largest absolute lift and the biggest exposure volume, and ChatGPT delivered the most consistent uplift across industries, sitting roughly 38 to 86% above baseline depending on the vertical.

Our view at Bushnote: this is the first time we've been able to point clients to hard evidence that being named by an AI is not a vanity metric. It changes behaviour. It moves people to your site. The mention is the media placement and the visit is the response. They're simply separated by hours and days, which is exactly why nobody could see the connection until now.

Why Your Dashboard Is Structurally Blind to This

Here's the part that reframes everything. According to Profound, even after ChatGPT's May 2026 update made its answers more clickable, only around 2.5% of these downstream visits carried any trackable AI-referral parameter. More than 97% of the traffic sparked by an AI mention arrived with no fingerprint at all.

This is not a tracking bug you can fix by bolting on another UTM. It is the nature of the channel. Someone reads a recommendation inside a conversation, sits with it, and returns later through a branded search or a direct visit. The moment of influence and the moment of the click have been pulled apart. Profound's framing is that AI visibility now behaves like a billboard. The effect shows up in what people do hours and days later, not in a neat, attributable click.

We think "billboard" is the right mental model, and it should be liberating rather than frustrating. No CMO has ever demanded a UTM parameter on a billboard over the M4. They accept that out-of-home advertising works through exposure and delayed response, and they measure it accordingly. AI visibility deserves the same maturity. The brands that keep insisting "if we can't tag it, it isn't real" are quietly defunding their fastest-growing discovery channel because their spreadsheet can't see it.

There's a distinctly Australian data point that sharpens this. According to Telsyte's 2026 research, 81% of Australians are aware that AI-generated summaries now appear in search results, and around half say they would often rely on that summary without clicking through to the original source. Read that twice. Half the country is prepared to act on an AI answer without ever visiting a website, which means the mention is the impression, and often the only one you get.

Why This Hits Harder in Australia Than Almost Anywhere

If AI mentions move markets, then the size of the exposed audience is the whole game, and Australia's audience is enormous relative to its size.

According to Roy Morgan's March-quarter 2026 research, 13.6 million Australians, or 58% of everyone aged 14 and over, used AI tools in an average four-week period, with ChatGPT alone reaching 10.5 million people, more than double its nearest rival. Telsyte's study, using a broader definition and age base, puts the figure higher again. According to Telsyte, 17.4 million Australians (77% of those aged 16 and over) now use AI, with the national user base having grown by more than six million people in a single year.

It is not just breadth, it is depth. According to Telsyte, 12% of Australians now name an AI tool as their primary way to find information online, up from just 5% a year earlier, more than doubling in twelve months. And this is not a youth phenomenon. According to Roy Morgan, the heaviest AI users are Australians aged 25 to 34 (74%) and 35 to 49 (72%), which is to say the working-age decision-makers and buyers most brands actually care about.

Then there's the signal we find most telling. According to Anthropic's Economic Index, Australia ranks first out of 121 countries for Claude usage per capita, running at more than four times, and by Anthropic's May 2026 update 6.4 times, the level the country's working-age population would predict. Australians don't just dabble in AI. On a per-person basis we are among the most intensive AI users on earth.

Our read at Bushnote: Australia is not a follower market waiting to see how AI search plays out overseas. We are a leading indicator. Whatever behavioural shift the Profound study captures in aggregate, it is landing here first and hardest. For an Australian brand, treating AI visibility as a "next year" problem is a competitive gift to whoever in your category treats it as a this-quarter one.

Not Every Platform Earns You the Same Return

One of the most practical things in Profound's data is that the uplift is not evenly distributed. According to Profound, the mention-to-visit lift varies considerably by industry and by platform, which means spreading your effort equally across every AI assistant is a mistake.

According to Profound's industry breakdown, the seven-day site-visit lift looked roughly like this:

The practical reading, in our words: the platform that best moves your category is not necessarily the one that moves the category next door. A retailer and a software business chasing the same "be everywhere" AI strategy will both waste effort. The retailer's biggest relative prize sits with Gemini. The software business gets outsized lift from Google AI Overviews. This is exactly the kind of allocation decision that separates a considered strategy from a scattergun one, and it's where we spend a lot of our time.

The Seven-Day Window Is the Honest Measure

Attribution debates usually collapse into a fight over the last click. Profound's timing data explains why that fight is the wrong one for AI.

According to Profound, only about 20.5% of first downstream visits happen within an hour of the AI mention, and 42% within 24 hours, which means the majority arrive after the first day. If you only measure same-session clicks, you are structurally throwing away most of the effect. A seven-day window, not a same-hour click, is the honest way to capture what an AI mention actually does.

For us this settles a long-running argument. When a client asks "how do we measure AI?", the answer is not a single referral number. It is three layers, measured together: visibility (do you appear in the answer at all), behaviour (does that appearance move people to act), and context (is the lift strong for your specific category and the specific platform your buyers use). Look at only one and you'll either over-invest in a weak channel or dismiss a strong one. Look at all three and the picture finally makes sense.

What We Tell Australian Brands to Do About It

None of this is theoretical for the businesses we work with, so here's the practical version.

Build to Be Mentioned, Not Just to Rank

Traditional SEO gets you into the source pool. It does not guarantee you're the name the model reaches for. Earning mentions means clear, credible, well-structured content that states what you do, who you serve and why you're a legitimate answer, the kind of material an AI can quote with confidence. This is the core of generative engine optimisation, and it is now as important as ranking.

Match Your Effort to the Platform That Moves Your Category

Use the Profound-style logic above. Don't spread thin. Put weight where your industry shows the strongest mention-to-visit lift, and treat the rest as maintenance.

Measure Over a Week, Not a Click

Stop grading AI visibility on same-session referrals. Watch branded search, direct traffic and assisted conversions across the seven days after your visibility improves. When the tagged number is near zero but branded demand climbs, that's the mention effect, not noise.

Protect the Trust That Earns the Click

According to Telsyte, only 27% of Australians are comfortable seeing ads inside AI-generated answers and 52% consider sponsored AI responses less trustworthy than organic ones. As AI platforms move toward advertising, being organically cited, named because you're a credible answer rather than because you paid, becomes a genuine trust advantage. That is worth defending.

The Bottom Line

AI is already sending Australians to your website. It just is not telling your analytics it did. The Profound study confirms the mechanism, the Australian adoption data confirms the scale, and our own client experience confirms the stakes. The brands that win from here won't be the ones with the cleanest attribution report, because that report will keep lying to them. They'll be the ones who accepted that AI visibility works like a billboard, built deliberately to be the name the model mentions, and measured the week that followed instead of the click that never came.

At Bushnote, that is the shift we're helping brands make. Quietly, and before their competitors notice. Full credit to Nikolas Laskaris and the Profound team for putting hard numbers to an effect the rest of the industry could only describe.